用awk递归函数

Question

Given a csv describing firstname and lastname of parent-child relationship

$ cat /var/tmp/hier
F2 L2,F1 L1
F3 L3,F1 L1
F4 L4,F2 L2
F5 L5,F2 L2
F6 L6,F3 L3

I want to print:

F1 L1
    F2 L2
        F4 L4
        F5 L5
    F3 L3
        F6 L6

I wrote a script like below:

#!/bin/bash
print_node() {
        echo "awk -F, '\$2=="\"$@\"" {print \$1}' /var/tmp/hier"
        for node in `eval "awk -F, '\$2=="\"$@\"" {print \$1}'     /var/tmp/hier"`
        do
                echo -e "\t"$node
                print_node "$node"
        done
}
print_node "$1"

run the script:

$ ./print_tree.sh "F1 L1"
awk -F, '$2=="F1 L1" {print $1}' /var/tmp/hier
awk: syntax error near line 1
awk: bailing out near line 1

It seemed that the awk command was malformed. but if I run the command shown in the debug output, it works:

$ awk -F, '$2=="F1 L1" {print $1}' /var/tmp/hier
F2 L2
F3 L3

Could anyone point out what is causing this error?

=======================
@moderatoors: I added perl and python tags because I'd welcome solutions in perl or python code. take no offence, please.

Answer 1

With GNU awk for multi-dimensional arrays:

$ cat tst.awk
BEGIN { FS="," }
function descend(node) {
    printf "%*s%s\n", indent, "", node
    if ( isarray(map[node]) ) {
        indent += 3
        for (child in map[node]) {
            descend(child)
        }
        indent -= 3
    }
    return
}
NR==1 { root = $2 }
{ map[$2][$1] }
END { descend(root) }

$ awk -f tst.awk file
F1 L1
   F2 L2
      F4 L4
      F5 L5
   F3 L3
      F6 L6

Answer 2

I would personally reach for Perl here; you could also do Python (or any other similar-level language that happens to be there, like Ruby or Tcl, but Perl and Python are almost universally preinstalled). I would use one of them since they have built-in nested data structures, which make it easy to cache the tree in navigable form, instead of re-parsing the parent links every time you want to fetch a node's children. (GNU awk has arrays of arrays, but BSD awk doesn't.)

Anyway, here's one perl solution:

#!/usr/bin/env perl
use strict;
use warnings;

my %parent;

while (<>) {
  chomp;
  my ($child, $parent) = split ',';
  $parent{$child} = $parent;
}

my (%children, %roots);

while (my ($child, $parent) = each %parent) {
  push @{$children{$parent} ||= []}, $child;
  $roots{$parent} = 1 unless $parent{$parent};
}

foreach my $root (sort keys %roots) {
  show($root);
}

sub show {
  my ($node, $indent) = (@_,'');
  print "$indent$node\n";
  foreach my $child (sort(@{$children{$node}||[]})) {
    show($child, "    $indent");
  }
}

I saved the above as print_tree.pl and ran it like this on your data:

$ perl print_tree.pl *csv

You could also make it executable with chmod +x print_tree.pl and run it without explicitly calling perl :

$ ./print_tree.pl *csv

Anyway, on your sample data, it produces this output:

F1 L1
    F2 L2
        F4 L4
        F5 L5
    F3 L3
        F6 L6

Answer 3

alternative solution without multi dimensional awk arrays which works for this hierarchy

join -t, -1 1 -2 2 inputfile{,} | awk -F, -f tree.awk

and awk script is as follows

$ cat tree.awk 
    {
            s=$1;$1=$2;$2=s;
            t=""
            for (i=1;i<=NF;i++) {
                    if (! ($i in n)) {
                            print t $i
                            n[$i]
                    }
                    t=t "\t"
            }
    }